af_inet.c 44 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858
  1. /*
  2. * INET An implementation of the TCP/IP protocol suite for the LINUX
  3. * operating system. INET is implemented using the BSD Socket
  4. * interface as the means of communication with the user level.
  5. *
  6. * PF_INET protocol family socket handler.
  7. *
  8. * Authors: Ross Biro
  9. * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  10. * Florian La Roche, <flla@stud.uni-sb.de>
  11. * Alan Cox, <A.Cox@swansea.ac.uk>
  12. *
  13. * Changes (see also sock.c)
  14. *
  15. * piggy,
  16. * Karl Knutson : Socket protocol table
  17. * A.N.Kuznetsov : Socket death error in accept().
  18. * John Richardson : Fix non blocking error in connect()
  19. * so sockets that fail to connect
  20. * don't return -EINPROGRESS.
  21. * Alan Cox : Asynchronous I/O support
  22. * Alan Cox : Keep correct socket pointer on sock
  23. * structures
  24. * when accept() ed
  25. * Alan Cox : Semantics of SO_LINGER aren't state
  26. * moved to close when you look carefully.
  27. * With this fixed and the accept bug fixed
  28. * some RPC stuff seems happier.
  29. * Niibe Yutaka : 4.4BSD style write async I/O
  30. * Alan Cox,
  31. * Tony Gale : Fixed reuse semantics.
  32. * Alan Cox : bind() shouldn't abort existing but dead
  33. * sockets. Stops FTP netin:.. I hope.
  34. * Alan Cox : bind() works correctly for RAW sockets.
  35. * Note that FreeBSD at least was broken
  36. * in this respect so be careful with
  37. * compatibility tests...
  38. * Alan Cox : routing cache support
  39. * Alan Cox : memzero the socket structure for
  40. * compactness.
  41. * Matt Day : nonblock connect error handler
  42. * Alan Cox : Allow large numbers of pending sockets
  43. * (eg for big web sites), but only if
  44. * specifically application requested.
  45. * Alan Cox : New buffering throughout IP. Used
  46. * dumbly.
  47. * Alan Cox : New buffering now used smartly.
  48. * Alan Cox : BSD rather than common sense
  49. * interpretation of listen.
  50. * Germano Caronni : Assorted small races.
  51. * Alan Cox : sendmsg/recvmsg basic support.
  52. * Alan Cox : Only sendmsg/recvmsg now supported.
  53. * Alan Cox : Locked down bind (see security list).
  54. * Alan Cox : Loosened bind a little.
  55. * Mike McLagan : ADD/DEL DLCI Ioctls
  56. * Willy Konynenberg : Transparent proxying support.
  57. * David S. Miller : New socket lookup architecture.
  58. * Some other random speedups.
  59. * Cyrus Durgin : Cleaned up file for kmod hacks.
  60. * Andi Kleen : Fix inet_stream_connect TCP race.
  61. *
  62. * This program is free software; you can redistribute it and/or
  63. * modify it under the terms of the GNU General Public License
  64. * as published by the Free Software Foundation; either version
  65. * 2 of the License, or (at your option) any later version.
  66. */
  67. #define pr_fmt(fmt) "IPv4: " fmt
  68. #include <linux/err.h>
  69. #include <linux/errno.h>
  70. #include <linux/types.h>
  71. #include <linux/socket.h>
  72. #include <linux/in.h>
  73. #include <linux/kernel.h>
  74. #include <linux/module.h>
  75. #include <linux/sched.h>
  76. #include <linux/timer.h>
  77. #include <linux/string.h>
  78. #include <linux/sockios.h>
  79. #include <linux/net.h>
  80. #include <linux/capability.h>
  81. #include <linux/fcntl.h>
  82. #include <linux/mm.h>
  83. #include <linux/interrupt.h>
  84. #include <linux/stat.h>
  85. #include <linux/init.h>
  86. #include <linux/poll.h>
  87. #include <linux/netfilter_ipv4.h>
  88. #include <linux/random.h>
  89. #include <linux/slab.h>
  90. #include <asm/uaccess.h>
  91. #include <linux/inet.h>
  92. #include <linux/igmp.h>
  93. #include <linux/inetdevice.h>
  94. #include <linux/netdevice.h>
  95. #include <net/checksum.h>
  96. #include <net/ip.h>
  97. #include <net/protocol.h>
  98. #include <net/arp.h>
  99. #include <net/route.h>
  100. #include <net/ip_fib.h>
  101. #include <net/inet_connection_sock.h>
  102. #include <net/tcp.h>
  103. #include <net/udp.h>
  104. #include <net/udplite.h>
  105. #include <net/ping.h>
  106. #include <linux/skbuff.h>
  107. #include <net/sock.h>
  108. #include <net/raw.h>
  109. #include <net/icmp.h>
  110. #include <net/ipip.h>
  111. #include <net/inet_common.h>
  112. #include <net/xfrm.h>
  113. #include <net/net_namespace.h>
  114. #ifdef CONFIG_IP_MROUTE
  115. #include <linux/mroute.h>
  116. #endif
  117. /* The inetsw table contains everything that inet_create needs to
  118. * build a new socket.
  119. */
  120. static struct list_head inetsw[SOCK_MAX];
  121. static DEFINE_SPINLOCK(inetsw_lock);
  122. struct ipv4_config ipv4_config;
  123. EXPORT_SYMBOL(ipv4_config);
  124. /* New destruction routine */
  125. void inet_sock_destruct(struct sock *sk)
  126. {
  127. struct inet_sock *inet = inet_sk(sk);
  128. __skb_queue_purge(&sk->sk_receive_queue);
  129. __skb_queue_purge(&sk->sk_error_queue);
  130. sk_mem_reclaim(sk);
  131. if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
  132. pr_err("Attempt to release TCP socket in state %d %p\n",
  133. sk->sk_state, sk);
  134. return;
  135. }
  136. if (!sock_flag(sk, SOCK_DEAD)) {
  137. pr_err("Attempt to release alive inet socket %p\n", sk);
  138. return;
  139. }
  140. WARN_ON(atomic_read(&sk->sk_rmem_alloc));
  141. WARN_ON(atomic_read(&sk->sk_wmem_alloc));
  142. WARN_ON(sk->sk_wmem_queued);
  143. WARN_ON(sk->sk_forward_alloc);
  144. kfree(rcu_dereference_protected(inet->inet_opt, 1));
  145. dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
  146. dst_release(sk->sk_rx_dst);
  147. sk_refcnt_debug_dec(sk);
  148. }
  149. EXPORT_SYMBOL(inet_sock_destruct);
  150. /*
  151. * The routines beyond this point handle the behaviour of an AF_INET
  152. * socket object. Mostly it punts to the subprotocols of IP to do
  153. * the work.
  154. */
  155. /*
  156. * Automatically bind an unbound socket.
  157. */
  158. static int inet_autobind(struct sock *sk)
  159. {
  160. struct inet_sock *inet;
  161. /* We may need to bind the socket. */
  162. lock_sock(sk);
  163. inet = inet_sk(sk);
  164. if (!inet->inet_num) {
  165. if (sk->sk_prot->get_port(sk, 0)) {
  166. release_sock(sk);
  167. return -EAGAIN;
  168. }
  169. inet->inet_sport = htons(inet->inet_num);
  170. }
  171. release_sock(sk);
  172. return 0;
  173. }
  174. /*
  175. * Move a socket into listening state.
  176. */
  177. int inet_listen(struct socket *sock, int backlog)
  178. {
  179. struct sock *sk = sock->sk;
  180. unsigned char old_state;
  181. int err;
  182. lock_sock(sk);
  183. err = -EINVAL;
  184. if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
  185. goto out;
  186. old_state = sk->sk_state;
  187. if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
  188. goto out;
  189. /* Really, if the socket is already in listen state
  190. * we can only allow the backlog to be adjusted.
  191. */
  192. if (old_state != TCP_LISTEN) {
  193. /* Check special setups for testing purpose to enable TFO w/o
  194. * requiring TCP_FASTOPEN sockopt.
  195. * Note that only TCP sockets (SOCK_STREAM) will reach here.
  196. * Also fastopenq may already been allocated because this
  197. * socket was in TCP_LISTEN state previously but was
  198. * shutdown() (rather than close()).
  199. */
  200. if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 &&
  201. inet_csk(sk)->icsk_accept_queue.fastopenq == NULL) {
  202. if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0)
  203. err = fastopen_init_queue(sk, backlog);
  204. else if ((sysctl_tcp_fastopen &
  205. TFO_SERVER_WO_SOCKOPT2) != 0)
  206. err = fastopen_init_queue(sk,
  207. ((uint)sysctl_tcp_fastopen) >> 16);
  208. else
  209. err = 0;
  210. if (err)
  211. goto out;
  212. }
  213. err = inet_csk_listen_start(sk, backlog);
  214. if (err)
  215. goto out;
  216. }
  217. sk->sk_max_ack_backlog = backlog;
  218. err = 0;
  219. out:
  220. release_sock(sk);
  221. return err;
  222. }
  223. EXPORT_SYMBOL(inet_listen);
  224. u32 inet_ehash_secret __read_mostly;
  225. EXPORT_SYMBOL(inet_ehash_secret);
  226. /*
  227. * inet_ehash_secret must be set exactly once
  228. */
  229. void build_ehash_secret(void)
  230. {
  231. u32 rnd;
  232. do {
  233. get_random_bytes(&rnd, sizeof(rnd));
  234. } while (rnd == 0);
  235. cmpxchg(&inet_ehash_secret, 0, rnd);
  236. }
  237. EXPORT_SYMBOL(build_ehash_secret);
  238. /*
  239. * Create an inet socket.
  240. */
  241. static int inet_create(struct net *net, struct socket *sock, int protocol,
  242. int kern)
  243. {
  244. struct sock *sk;
  245. struct inet_protosw *answer;
  246. struct inet_sock *inet;
  247. struct proto *answer_prot;
  248. unsigned char answer_flags;
  249. char answer_no_check;
  250. int try_loading_module = 0;
  251. int err;
  252. if (unlikely(!inet_ehash_secret))
  253. if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
  254. build_ehash_secret();
  255. sock->state = SS_UNCONNECTED;
  256. /* Look for the requested type/protocol pair. */
  257. lookup_protocol:
  258. err = -ESOCKTNOSUPPORT;
  259. rcu_read_lock();
  260. list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
  261. err = 0;
  262. /* Check the non-wild match. */
  263. if (protocol == answer->protocol) {
  264. if (protocol != IPPROTO_IP)
  265. break;
  266. } else {
  267. /* Check for the two wild cases. */
  268. if (IPPROTO_IP == protocol) {
  269. protocol = answer->protocol;
  270. break;
  271. }
  272. if (IPPROTO_IP == answer->protocol)
  273. break;
  274. }
  275. err = -EPROTONOSUPPORT;
  276. }
  277. if (unlikely(err)) {
  278. if (try_loading_module < 2) {
  279. rcu_read_unlock();
  280. /*
  281. * Be more specific, e.g. net-pf-2-proto-132-type-1
  282. * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
  283. */
  284. if (++try_loading_module == 1)
  285. request_module("net-pf-%d-proto-%d-type-%d",
  286. PF_INET, protocol, sock->type);
  287. /*
  288. * Fall back to generic, e.g. net-pf-2-proto-132
  289. * (net-pf-PF_INET-proto-IPPROTO_SCTP)
  290. */
  291. else
  292. request_module("net-pf-%d-proto-%d",
  293. PF_INET, protocol);
  294. goto lookup_protocol;
  295. } else
  296. goto out_rcu_unlock;
  297. }
  298. err = -EPERM;
  299. if (sock->type == SOCK_RAW && !kern &&
  300. !ns_capable(net->user_ns, CAP_NET_RAW))
  301. goto out_rcu_unlock;
  302. sock->ops = answer->ops;
  303. answer_prot = answer->prot;
  304. answer_no_check = answer->no_check;
  305. answer_flags = answer->flags;
  306. rcu_read_unlock();
  307. WARN_ON(answer_prot->slab == NULL);
  308. err = -ENOBUFS;
  309. sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
  310. if (sk == NULL)
  311. goto out;
  312. err = 0;
  313. sk->sk_no_check = answer_no_check;
  314. if (INET_PROTOSW_REUSE & answer_flags)
  315. sk->sk_reuse = SK_CAN_REUSE;
  316. inet = inet_sk(sk);
  317. inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
  318. inet->nodefrag = 0;
  319. if (SOCK_RAW == sock->type) {
  320. inet->inet_num = protocol;
  321. if (IPPROTO_RAW == protocol)
  322. inet->hdrincl = 1;
  323. }
  324. if (ipv4_config.no_pmtu_disc)
  325. inet->pmtudisc = IP_PMTUDISC_DONT;
  326. else
  327. inet->pmtudisc = IP_PMTUDISC_WANT;
  328. inet->inet_id = 0;
  329. sock_init_data(sock, sk);
  330. sk->sk_destruct = inet_sock_destruct;
  331. sk->sk_protocol = protocol;
  332. sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
  333. inet->uc_ttl = -1;
  334. inet->mc_loop = 1;
  335. inet->mc_ttl = 1;
  336. inet->mc_all = 1;
  337. inet->mc_index = 0;
  338. inet->mc_list = NULL;
  339. inet->rcv_tos = 0;
  340. sk_refcnt_debug_inc(sk);
  341. if (inet->inet_num) {
  342. /* It assumes that any protocol which allows
  343. * the user to assign a number at socket
  344. * creation time automatically
  345. * shares.
  346. */
  347. inet->inet_sport = htons(inet->inet_num);
  348. /* Add to protocol hash chains. */
  349. sk->sk_prot->hash(sk);
  350. }
  351. if (sk->sk_prot->init) {
  352. err = sk->sk_prot->init(sk);
  353. if (err)
  354. sk_common_release(sk);
  355. }
  356. out:
  357. return err;
  358. out_rcu_unlock:
  359. rcu_read_unlock();
  360. goto out;
  361. }
  362. /*
  363. * The peer socket should always be NULL (or else). When we call this
  364. * function we are destroying the object and from then on nobody
  365. * should refer to it.
  366. */
  367. int inet_release(struct socket *sock)
  368. {
  369. struct sock *sk = sock->sk;
  370. if (sk) {
  371. long timeout;
  372. sock_rps_reset_flow(sk);
  373. /* Applications forget to leave groups before exiting */
  374. ip_mc_drop_socket(sk);
  375. /* If linger is set, we don't return until the close
  376. * is complete. Otherwise we return immediately. The
  377. * actually closing is done the same either way.
  378. *
  379. * If the close is due to the process exiting, we never
  380. * linger..
  381. */
  382. timeout = 0;
  383. if (sock_flag(sk, SOCK_LINGER) &&
  384. !(current->flags & PF_EXITING))
  385. timeout = sk->sk_lingertime;
  386. sock->sk = NULL;
  387. sk->sk_prot->close(sk, timeout);
  388. }
  389. return 0;
  390. }
  391. EXPORT_SYMBOL(inet_release);
  392. /* It is off by default, see below. */
  393. int sysctl_ip_nonlocal_bind __read_mostly;
  394. EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
  395. int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
  396. {
  397. struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
  398. struct sock *sk = sock->sk;
  399. struct inet_sock *inet = inet_sk(sk);
  400. struct net *net = sock_net(sk);
  401. unsigned short snum;
  402. int chk_addr_ret;
  403. int err;
  404. /* If the socket has its own bind function then use it. (RAW) */
  405. if (sk->sk_prot->bind) {
  406. err = sk->sk_prot->bind(sk, uaddr, addr_len);
  407. goto out;
  408. }
  409. err = -EINVAL;
  410. if (addr_len < sizeof(struct sockaddr_in))
  411. goto out;
  412. if (addr->sin_family != AF_INET) {
  413. /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
  414. * only if s_addr is INADDR_ANY.
  415. */
  416. err = -EAFNOSUPPORT;
  417. if (addr->sin_family != AF_UNSPEC ||
  418. addr->sin_addr.s_addr != htonl(INADDR_ANY))
  419. goto out;
  420. }
  421. chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
  422. /* Not specified by any standard per-se, however it breaks too
  423. * many applications when removed. It is unfortunate since
  424. * allowing applications to make a non-local bind solves
  425. * several problems with systems using dynamic addressing.
  426. * (ie. your servers still start up even if your ISDN link
  427. * is temporarily down)
  428. */
  429. err = -EADDRNOTAVAIL;
  430. if (!sysctl_ip_nonlocal_bind &&
  431. !(inet->freebind || inet->transparent) &&
  432. addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
  433. chk_addr_ret != RTN_LOCAL &&
  434. chk_addr_ret != RTN_MULTICAST &&
  435. chk_addr_ret != RTN_BROADCAST)
  436. goto out;
  437. snum = ntohs(addr->sin_port);
  438. err = -EACCES;
  439. if (snum && snum < PROT_SOCK &&
  440. !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
  441. goto out;
  442. /* We keep a pair of addresses. rcv_saddr is the one
  443. * used by hash lookups, and saddr is used for transmit.
  444. *
  445. * In the BSD API these are the same except where it
  446. * would be illegal to use them (multicast/broadcast) in
  447. * which case the sending device address is used.
  448. */
  449. lock_sock(sk);
  450. /* Check these errors (active socket, double bind). */
  451. err = -EINVAL;
  452. if (sk->sk_state != TCP_CLOSE || inet->inet_num)
  453. goto out_release_sock;
  454. inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
  455. if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
  456. inet->inet_saddr = 0; /* Use device */
  457. /* Make sure we are allowed to bind here. */
  458. if (sk->sk_prot->get_port(sk, snum)) {
  459. inet->inet_saddr = inet->inet_rcv_saddr = 0;
  460. err = -EADDRINUSE;
  461. goto out_release_sock;
  462. }
  463. if (inet->inet_rcv_saddr)
  464. sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
  465. if (snum)
  466. sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
  467. inet->inet_sport = htons(inet->inet_num);
  468. inet->inet_daddr = 0;
  469. inet->inet_dport = 0;
  470. sk_dst_reset(sk);
  471. err = 0;
  472. out_release_sock:
  473. release_sock(sk);
  474. out:
  475. return err;
  476. }
  477. EXPORT_SYMBOL(inet_bind);
  478. int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
  479. int addr_len, int flags)
  480. {
  481. struct sock *sk = sock->sk;
  482. if (addr_len < sizeof(uaddr->sa_family))
  483. return -EINVAL;
  484. if (uaddr->sa_family == AF_UNSPEC)
  485. return sk->sk_prot->disconnect(sk, flags);
  486. if (!inet_sk(sk)->inet_num && inet_autobind(sk))
  487. return -EAGAIN;
  488. return sk->sk_prot->connect(sk, uaddr, addr_len);
  489. }
  490. EXPORT_SYMBOL(inet_dgram_connect);
  491. static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
  492. {
  493. DEFINE_WAIT(wait);
  494. prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
  495. sk->sk_write_pending += writebias;
  496. /* Basic assumption: if someone sets sk->sk_err, he _must_
  497. * change state of the socket from TCP_SYN_*.
  498. * Connect() does not allow to get error notifications
  499. * without closing the socket.
  500. */
  501. while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
  502. release_sock(sk);
  503. timeo = schedule_timeout(timeo);
  504. lock_sock(sk);
  505. if (signal_pending(current) || !timeo)
  506. break;
  507. prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
  508. }
  509. finish_wait(sk_sleep(sk), &wait);
  510. sk->sk_write_pending -= writebias;
  511. return timeo;
  512. }
  513. /*
  514. * Connect to a remote host. There is regrettably still a little
  515. * TCP 'magic' in here.
  516. */
  517. int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
  518. int addr_len, int flags)
  519. {
  520. struct sock *sk = sock->sk;
  521. int err;
  522. long timeo;
  523. if (addr_len < sizeof(uaddr->sa_family))
  524. return -EINVAL;
  525. if (uaddr->sa_family == AF_UNSPEC) {
  526. err = sk->sk_prot->disconnect(sk, flags);
  527. sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
  528. goto out;
  529. }
  530. switch (sock->state) {
  531. default:
  532. err = -EINVAL;
  533. goto out;
  534. case SS_CONNECTED:
  535. err = -EISCONN;
  536. goto out;
  537. case SS_CONNECTING:
  538. err = -EALREADY;
  539. /* Fall out of switch with err, set for this state */
  540. break;
  541. case SS_UNCONNECTED:
  542. err = -EISCONN;
  543. if (sk->sk_state != TCP_CLOSE)
  544. goto out;
  545. err = sk->sk_prot->connect(sk, uaddr, addr_len);
  546. if (err < 0)
  547. goto out;
  548. sock->state = SS_CONNECTING;
  549. /* Just entered SS_CONNECTING state; the only
  550. * difference is that return value in non-blocking
  551. * case is EINPROGRESS, rather than EALREADY.
  552. */
  553. err = -EINPROGRESS;
  554. break;
  555. }
  556. timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
  557. if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
  558. int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
  559. tcp_sk(sk)->fastopen_req &&
  560. tcp_sk(sk)->fastopen_req->data ? 1 : 0;
  561. /* Error code is set above */
  562. if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
  563. goto out;
  564. err = sock_intr_errno(timeo);
  565. if (signal_pending(current))
  566. goto out;
  567. }
  568. /* Connection was closed by RST, timeout, ICMP error
  569. * or another process disconnected us.
  570. */
  571. if (sk->sk_state == TCP_CLOSE)
  572. goto sock_error;
  573. /* sk->sk_err may be not zero now, if RECVERR was ordered by user
  574. * and error was received after socket entered established state.
  575. * Hence, it is handled normally after connect() return successfully.
  576. */
  577. sock->state = SS_CONNECTED;
  578. err = 0;
  579. out:
  580. return err;
  581. sock_error:
  582. err = sock_error(sk) ? : -ECONNABORTED;
  583. sock->state = SS_UNCONNECTED;
  584. if (sk->sk_prot->disconnect(sk, flags))
  585. sock->state = SS_DISCONNECTING;
  586. goto out;
  587. }
  588. EXPORT_SYMBOL(__inet_stream_connect);
  589. int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
  590. int addr_len, int flags)
  591. {
  592. int err;
  593. lock_sock(sock->sk);
  594. err = __inet_stream_connect(sock, uaddr, addr_len, flags);
  595. release_sock(sock->sk);
  596. return err;
  597. }
  598. EXPORT_SYMBOL(inet_stream_connect);
  599. /*
  600. * Accept a pending connection. The TCP layer now gives BSD semantics.
  601. */
  602. int inet_accept(struct socket *sock, struct socket *newsock, int flags)
  603. {
  604. struct sock *sk1 = sock->sk;
  605. int err = -EINVAL;
  606. struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
  607. if (!sk2)
  608. goto do_err;
  609. lock_sock(sk2);
  610. sock_rps_record_flow(sk2);
  611. WARN_ON(!((1 << sk2->sk_state) &
  612. (TCPF_ESTABLISHED | TCPF_SYN_RECV |
  613. TCPF_CLOSE_WAIT | TCPF_CLOSE)));
  614. sock_graft(sk2, newsock);
  615. newsock->state = SS_CONNECTED;
  616. err = 0;
  617. release_sock(sk2);
  618. do_err:
  619. return err;
  620. }
  621. EXPORT_SYMBOL(inet_accept);
  622. /*
  623. * This does both peername and sockname.
  624. */
  625. int inet_getname(struct socket *sock, struct sockaddr *uaddr,
  626. int *uaddr_len, int peer)
  627. {
  628. struct sock *sk = sock->sk;
  629. struct inet_sock *inet = inet_sk(sk);
  630. DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
  631. sin->sin_family = AF_INET;
  632. if (peer) {
  633. if (!inet->inet_dport ||
  634. (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
  635. peer == 1))
  636. return -ENOTCONN;
  637. sin->sin_port = inet->inet_dport;
  638. sin->sin_addr.s_addr = inet->inet_daddr;
  639. } else {
  640. __be32 addr = inet->inet_rcv_saddr;
  641. if (!addr)
  642. addr = inet->inet_saddr;
  643. sin->sin_port = inet->inet_sport;
  644. sin->sin_addr.s_addr = addr;
  645. }
  646. memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
  647. *uaddr_len = sizeof(*sin);
  648. return 0;
  649. }
  650. EXPORT_SYMBOL(inet_getname);
  651. int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
  652. size_t size)
  653. {
  654. struct sock *sk = sock->sk;
  655. sock_rps_record_flow(sk);
  656. /* We may need to bind the socket. */
  657. if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
  658. inet_autobind(sk))
  659. return -EAGAIN;
  660. return sk->sk_prot->sendmsg(iocb, sk, msg, size);
  661. }
  662. EXPORT_SYMBOL(inet_sendmsg);
  663. ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
  664. size_t size, int flags)
  665. {
  666. struct sock *sk = sock->sk;
  667. sock_rps_record_flow(sk);
  668. /* We may need to bind the socket. */
  669. if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
  670. inet_autobind(sk))
  671. return -EAGAIN;
  672. if (sk->sk_prot->sendpage)
  673. return sk->sk_prot->sendpage(sk, page, offset, size, flags);
  674. return sock_no_sendpage(sock, page, offset, size, flags);
  675. }
  676. EXPORT_SYMBOL(inet_sendpage);
  677. int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
  678. size_t size, int flags)
  679. {
  680. struct sock *sk = sock->sk;
  681. int addr_len = 0;
  682. int err;
  683. sock_rps_record_flow(sk);
  684. err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
  685. flags & ~MSG_DONTWAIT, &addr_len);
  686. if (err >= 0)
  687. msg->msg_namelen = addr_len;
  688. return err;
  689. }
  690. EXPORT_SYMBOL(inet_recvmsg);
  691. int inet_shutdown(struct socket *sock, int how)
  692. {
  693. struct sock *sk = sock->sk;
  694. int err = 0;
  695. /* This should really check to make sure
  696. * the socket is a TCP socket. (WHY AC...)
  697. */
  698. how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
  699. 1->2 bit 2 snds.
  700. 2->3 */
  701. if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
  702. return -EINVAL;
  703. lock_sock(sk);
  704. if (sock->state == SS_CONNECTING) {
  705. if ((1 << sk->sk_state) &
  706. (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
  707. sock->state = SS_DISCONNECTING;
  708. else
  709. sock->state = SS_CONNECTED;
  710. }
  711. switch (sk->sk_state) {
  712. case TCP_CLOSE:
  713. err = -ENOTCONN;
  714. /* Hack to wake up other listeners, who can poll for
  715. POLLHUP, even on eg. unconnected UDP sockets -- RR */
  716. default:
  717. sk->sk_shutdown |= how;
  718. if (sk->sk_prot->shutdown)
  719. sk->sk_prot->shutdown(sk, how);
  720. break;
  721. /* Remaining two branches are temporary solution for missing
  722. * close() in multithreaded environment. It is _not_ a good idea,
  723. * but we have no choice until close() is repaired at VFS level.
  724. */
  725. case TCP_LISTEN:
  726. if (!(how & RCV_SHUTDOWN))
  727. break;
  728. /* Fall through */
  729. case TCP_SYN_SENT:
  730. err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
  731. sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
  732. break;
  733. }
  734. /* Wake up anyone sleeping in poll. */
  735. sk->sk_state_change(sk);
  736. release_sock(sk);
  737. return err;
  738. }
  739. EXPORT_SYMBOL(inet_shutdown);
  740. /*
  741. * ioctl() calls you can issue on an INET socket. Most of these are
  742. * device configuration and stuff and very rarely used. Some ioctls
  743. * pass on to the socket itself.
  744. *
  745. * NOTE: I like the idea of a module for the config stuff. ie ifconfig
  746. * loads the devconfigure module does its configuring and unloads it.
  747. * There's a good 20K of config code hanging around the kernel.
  748. */
  749. int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
  750. {
  751. struct sock *sk = sock->sk;
  752. int err = 0;
  753. struct net *net = sock_net(sk);
  754. switch (cmd) {
  755. case SIOCGSTAMP:
  756. err = sock_get_timestamp(sk, (struct timeval __user *)arg);
  757. break;
  758. case SIOCGSTAMPNS:
  759. err = sock_get_timestampns(sk, (struct timespec __user *)arg);
  760. break;
  761. case SIOCADDRT:
  762. case SIOCDELRT:
  763. case SIOCRTMSG:
  764. err = ip_rt_ioctl(net, cmd, (void __user *)arg);
  765. break;
  766. case SIOCDARP:
  767. case SIOCGARP:
  768. case SIOCSARP:
  769. err = arp_ioctl(net, cmd, (void __user *)arg);
  770. break;
  771. case SIOCGIFADDR:
  772. case SIOCSIFADDR:
  773. case SIOCGIFBRDADDR:
  774. case SIOCSIFBRDADDR:
  775. case SIOCGIFNETMASK:
  776. case SIOCSIFNETMASK:
  777. case SIOCGIFDSTADDR:
  778. case SIOCSIFDSTADDR:
  779. case SIOCSIFPFLAGS:
  780. case SIOCGIFPFLAGS:
  781. case SIOCSIFFLAGS:
  782. err = devinet_ioctl(net, cmd, (void __user *)arg);
  783. break;
  784. default:
  785. if (sk->sk_prot->ioctl)
  786. err = sk->sk_prot->ioctl(sk, cmd, arg);
  787. else
  788. err = -ENOIOCTLCMD;
  789. break;
  790. }
  791. return err;
  792. }
  793. EXPORT_SYMBOL(inet_ioctl);
  794. #ifdef CONFIG_COMPAT
  795. static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
  796. {
  797. struct sock *sk = sock->sk;
  798. int err = -ENOIOCTLCMD;
  799. if (sk->sk_prot->compat_ioctl)
  800. err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
  801. return err;
  802. }
  803. #endif
  804. const struct proto_ops inet_stream_ops = {
  805. .family = PF_INET,
  806. .owner = THIS_MODULE,
  807. .release = inet_release,
  808. .bind = inet_bind,
  809. .connect = inet_stream_connect,
  810. .socketpair = sock_no_socketpair,
  811. .accept = inet_accept,
  812. .getname = inet_getname,
  813. .poll = tcp_poll,
  814. .ioctl = inet_ioctl,
  815. .listen = inet_listen,
  816. .shutdown = inet_shutdown,
  817. .setsockopt = sock_common_setsockopt,
  818. .getsockopt = sock_common_getsockopt,
  819. .sendmsg = inet_sendmsg,
  820. .recvmsg = inet_recvmsg,
  821. .mmap = sock_no_mmap,
  822. .sendpage = inet_sendpage,
  823. .splice_read = tcp_splice_read,
  824. #ifdef CONFIG_COMPAT
  825. .compat_setsockopt = compat_sock_common_setsockopt,
  826. .compat_getsockopt = compat_sock_common_getsockopt,
  827. .compat_ioctl = inet_compat_ioctl,
  828. #endif
  829. };
  830. EXPORT_SYMBOL(inet_stream_ops);
  831. const struct proto_ops inet_dgram_ops = {
  832. .family = PF_INET,
  833. .owner = THIS_MODULE,
  834. .release = inet_release,
  835. .bind = inet_bind,
  836. .connect = inet_dgram_connect,
  837. .socketpair = sock_no_socketpair,
  838. .accept = sock_no_accept,
  839. .getname = inet_getname,
  840. .poll = udp_poll,
  841. .ioctl = inet_ioctl,
  842. .listen = sock_no_listen,
  843. .shutdown = inet_shutdown,
  844. .setsockopt = sock_common_setsockopt,
  845. .getsockopt = sock_common_getsockopt,
  846. .sendmsg = inet_sendmsg,
  847. .recvmsg = inet_recvmsg,
  848. .mmap = sock_no_mmap,
  849. .sendpage = inet_sendpage,
  850. #ifdef CONFIG_COMPAT
  851. .compat_setsockopt = compat_sock_common_setsockopt,
  852. .compat_getsockopt = compat_sock_common_getsockopt,
  853. .compat_ioctl = inet_compat_ioctl,
  854. #endif
  855. };
  856. EXPORT_SYMBOL(inet_dgram_ops);
  857. /*
  858. * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
  859. * udp_poll
  860. */
  861. static const struct proto_ops inet_sockraw_ops = {
  862. .family = PF_INET,
  863. .owner = THIS_MODULE,
  864. .release = inet_release,
  865. .bind = inet_bind,
  866. .connect = inet_dgram_connect,
  867. .socketpair = sock_no_socketpair,
  868. .accept = sock_no_accept,
  869. .getname = inet_getname,
  870. .poll = datagram_poll,
  871. .ioctl = inet_ioctl,
  872. .listen = sock_no_listen,
  873. .shutdown = inet_shutdown,
  874. .setsockopt = sock_common_setsockopt,
  875. .getsockopt = sock_common_getsockopt,
  876. .sendmsg = inet_sendmsg,
  877. .recvmsg = inet_recvmsg,
  878. .mmap = sock_no_mmap,
  879. .sendpage = inet_sendpage,
  880. #ifdef CONFIG_COMPAT
  881. .compat_setsockopt = compat_sock_common_setsockopt,
  882. .compat_getsockopt = compat_sock_common_getsockopt,
  883. .compat_ioctl = inet_compat_ioctl,
  884. #endif
  885. };
  886. static const struct net_proto_family inet_family_ops = {
  887. .family = PF_INET,
  888. .create = inet_create,
  889. .owner = THIS_MODULE,
  890. };
  891. /* Upon startup we insert all the elements in inetsw_array[] into
  892. * the linked list inetsw.
  893. */
  894. static struct inet_protosw inetsw_array[] =
  895. {
  896. {
  897. .type = SOCK_STREAM,
  898. .protocol = IPPROTO_TCP,
  899. .prot = &tcp_prot,
  900. .ops = &inet_stream_ops,
  901. .no_check = 0,
  902. .flags = INET_PROTOSW_PERMANENT |
  903. INET_PROTOSW_ICSK,
  904. },
  905. {
  906. .type = SOCK_DGRAM,
  907. .protocol = IPPROTO_UDP,
  908. .prot = &udp_prot,
  909. .ops = &inet_dgram_ops,
  910. .no_check = UDP_CSUM_DEFAULT,
  911. .flags = INET_PROTOSW_PERMANENT,
  912. },
  913. {
  914. .type = SOCK_DGRAM,
  915. .protocol = IPPROTO_ICMP,
  916. .prot = &ping_prot,
  917. .ops = &inet_dgram_ops,
  918. .no_check = UDP_CSUM_DEFAULT,
  919. .flags = INET_PROTOSW_REUSE,
  920. },
  921. {
  922. .type = SOCK_RAW,
  923. .protocol = IPPROTO_IP, /* wild card */
  924. .prot = &raw_prot,
  925. .ops = &inet_sockraw_ops,
  926. .no_check = UDP_CSUM_DEFAULT,
  927. .flags = INET_PROTOSW_REUSE,
  928. }
  929. };
  930. #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
  931. void inet_register_protosw(struct inet_protosw *p)
  932. {
  933. struct list_head *lh;
  934. struct inet_protosw *answer;
  935. int protocol = p->protocol;
  936. struct list_head *last_perm;
  937. spin_lock_bh(&inetsw_lock);
  938. if (p->type >= SOCK_MAX)
  939. goto out_illegal;
  940. /* If we are trying to override a permanent protocol, bail. */
  941. answer = NULL;
  942. last_perm = &inetsw[p->type];
  943. list_for_each(lh, &inetsw[p->type]) {
  944. answer = list_entry(lh, struct inet_protosw, list);
  945. /* Check only the non-wild match. */
  946. if (INET_PROTOSW_PERMANENT & answer->flags) {
  947. if (protocol == answer->protocol)
  948. break;
  949. last_perm = lh;
  950. }
  951. answer = NULL;
  952. }
  953. if (answer)
  954. goto out_permanent;
  955. /* Add the new entry after the last permanent entry if any, so that
  956. * the new entry does not override a permanent entry when matched with
  957. * a wild-card protocol. But it is allowed to override any existing
  958. * non-permanent entry. This means that when we remove this entry, the
  959. * system automatically returns to the old behavior.
  960. */
  961. list_add_rcu(&p->list, last_perm);
  962. out:
  963. spin_unlock_bh(&inetsw_lock);
  964. return;
  965. out_permanent:
  966. pr_err("Attempt to override permanent protocol %d\n", protocol);
  967. goto out;
  968. out_illegal:
  969. pr_err("Ignoring attempt to register invalid socket type %d\n",
  970. p->type);
  971. goto out;
  972. }
  973. EXPORT_SYMBOL(inet_register_protosw);
  974. void inet_unregister_protosw(struct inet_protosw *p)
  975. {
  976. if (INET_PROTOSW_PERMANENT & p->flags) {
  977. pr_err("Attempt to unregister permanent protocol %d\n",
  978. p->protocol);
  979. } else {
  980. spin_lock_bh(&inetsw_lock);
  981. list_del_rcu(&p->list);
  982. spin_unlock_bh(&inetsw_lock);
  983. synchronize_net();
  984. }
  985. }
  986. EXPORT_SYMBOL(inet_unregister_protosw);
  987. /*
  988. * Shall we try to damage output packets if routing dev changes?
  989. */
  990. int sysctl_ip_dynaddr __read_mostly;
  991. static int inet_sk_reselect_saddr(struct sock *sk)
  992. {
  993. struct inet_sock *inet = inet_sk(sk);
  994. __be32 old_saddr = inet->inet_saddr;
  995. __be32 daddr = inet->inet_daddr;
  996. struct flowi4 *fl4;
  997. struct rtable *rt;
  998. __be32 new_saddr;
  999. struct ip_options_rcu *inet_opt;
  1000. inet_opt = rcu_dereference_protected(inet->inet_opt,
  1001. sock_owned_by_user(sk));
  1002. if (inet_opt && inet_opt->opt.srr)
  1003. daddr = inet_opt->opt.faddr;
  1004. /* Query new route. */
  1005. fl4 = &inet->cork.fl.u.ip4;
  1006. rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
  1007. sk->sk_bound_dev_if, sk->sk_protocol,
  1008. inet->inet_sport, inet->inet_dport, sk, false);
  1009. if (IS_ERR(rt))
  1010. return PTR_ERR(rt);
  1011. sk_setup_caps(sk, &rt->dst);
  1012. new_saddr = fl4->saddr;
  1013. if (new_saddr == old_saddr)
  1014. return 0;
  1015. if (sysctl_ip_dynaddr > 1) {
  1016. pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
  1017. __func__, &old_saddr, &new_saddr);
  1018. }
  1019. inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
  1020. /*
  1021. * XXX The only one ugly spot where we need to
  1022. * XXX really change the sockets identity after
  1023. * XXX it has entered the hashes. -DaveM
  1024. *
  1025. * Besides that, it does not check for connection
  1026. * uniqueness. Wait for troubles.
  1027. */
  1028. __sk_prot_rehash(sk);
  1029. return 0;
  1030. }
  1031. int inet_sk_rebuild_header(struct sock *sk)
  1032. {
  1033. struct inet_sock *inet = inet_sk(sk);
  1034. struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
  1035. __be32 daddr;
  1036. struct ip_options_rcu *inet_opt;
  1037. struct flowi4 *fl4;
  1038. int err;
  1039. /* Route is OK, nothing to do. */
  1040. if (rt)
  1041. return 0;
  1042. /* Reroute. */
  1043. rcu_read_lock();
  1044. inet_opt = rcu_dereference(inet->inet_opt);
  1045. daddr = inet->inet_daddr;
  1046. if (inet_opt && inet_opt->opt.srr)
  1047. daddr = inet_opt->opt.faddr;
  1048. rcu_read_unlock();
  1049. fl4 = &inet->cork.fl.u.ip4;
  1050. rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
  1051. inet->inet_dport, inet->inet_sport,
  1052. sk->sk_protocol, RT_CONN_FLAGS(sk),
  1053. sk->sk_bound_dev_if);
  1054. if (!IS_ERR(rt)) {
  1055. err = 0;
  1056. sk_setup_caps(sk, &rt->dst);
  1057. } else {
  1058. err = PTR_ERR(rt);
  1059. /* Routing failed... */
  1060. sk->sk_route_caps = 0;
  1061. /*
  1062. * Other protocols have to map its equivalent state to TCP_SYN_SENT.
  1063. * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
  1064. */
  1065. if (!sysctl_ip_dynaddr ||
  1066. sk->sk_state != TCP_SYN_SENT ||
  1067. (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
  1068. (err = inet_sk_reselect_saddr(sk)) != 0)
  1069. sk->sk_err_soft = -err;
  1070. }
  1071. return err;
  1072. }
  1073. EXPORT_SYMBOL(inet_sk_rebuild_header);
  1074. static int inet_gso_send_check(struct sk_buff *skb)
  1075. {
  1076. const struct net_offload *ops;
  1077. const struct iphdr *iph;
  1078. int proto;
  1079. int ihl;
  1080. int err = -EINVAL;
  1081. if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
  1082. goto out;
  1083. iph = ip_hdr(skb);
  1084. ihl = iph->ihl * 4;
  1085. if (ihl < sizeof(*iph))
  1086. goto out;
  1087. if (unlikely(!pskb_may_pull(skb, ihl)))
  1088. goto out;
  1089. __skb_pull(skb, ihl);
  1090. skb_reset_transport_header(skb);
  1091. iph = ip_hdr(skb);
  1092. proto = iph->protocol;
  1093. err = -EPROTONOSUPPORT;
  1094. rcu_read_lock();
  1095. ops = rcu_dereference(inet_offloads[proto]);
  1096. if (likely(ops && ops->callbacks.gso_send_check))
  1097. err = ops->callbacks.gso_send_check(skb);
  1098. rcu_read_unlock();
  1099. out:
  1100. return err;
  1101. }
  1102. static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
  1103. netdev_features_t features)
  1104. {
  1105. struct sk_buff *segs = ERR_PTR(-EINVAL);
  1106. const struct net_offload *ops;
  1107. struct iphdr *iph;
  1108. int proto;
  1109. int ihl;
  1110. int id;
  1111. unsigned int offset = 0;
  1112. if (!(features & NETIF_F_V4_CSUM))
  1113. features &= ~NETIF_F_SG;
  1114. if (unlikely(skb_shinfo(skb)->gso_type &
  1115. ~(SKB_GSO_TCPV4 |
  1116. SKB_GSO_UDP |
  1117. SKB_GSO_DODGY |
  1118. SKB_GSO_TCP_ECN |
  1119. SKB_GSO_GRE |
  1120. 0)))
  1121. goto out;
  1122. if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
  1123. goto out;
  1124. iph = ip_hdr(skb);
  1125. ihl = iph->ihl * 4;
  1126. if (ihl < sizeof(*iph))
  1127. goto out;
  1128. if (unlikely(!pskb_may_pull(skb, ihl)))
  1129. goto out;
  1130. __skb_pull(skb, ihl);
  1131. skb_reset_transport_header(skb);
  1132. iph = ip_hdr(skb);
  1133. id = ntohs(iph->id);
  1134. proto = iph->protocol;
  1135. segs = ERR_PTR(-EPROTONOSUPPORT);
  1136. rcu_read_lock();
  1137. ops = rcu_dereference(inet_offloads[proto]);
  1138. if (likely(ops && ops->callbacks.gso_segment))
  1139. segs = ops->callbacks.gso_segment(skb, features);
  1140. rcu_read_unlock();
  1141. if (IS_ERR_OR_NULL(segs))
  1142. goto out;
  1143. skb = segs;
  1144. do {
  1145. iph = ip_hdr(skb);
  1146. if (proto == IPPROTO_UDP) {
  1147. iph->id = htons(id);
  1148. iph->frag_off = htons(offset >> 3);
  1149. if (skb->next != NULL)
  1150. iph->frag_off |= htons(IP_MF);
  1151. offset += (skb->len - skb->mac_len - iph->ihl * 4);
  1152. } else
  1153. iph->id = htons(id++);
  1154. iph->tot_len = htons(skb->len - skb->mac_len);
  1155. iph->check = 0;
  1156. iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
  1157. } while ((skb = skb->next));
  1158. out:
  1159. return segs;
  1160. }
  1161. static struct sk_buff **inet_gro_receive(struct sk_buff **head,
  1162. struct sk_buff *skb)
  1163. {
  1164. const struct net_offload *ops;
  1165. struct sk_buff **pp = NULL;
  1166. struct sk_buff *p;
  1167. const struct iphdr *iph;
  1168. unsigned int hlen;
  1169. unsigned int off;
  1170. unsigned int id;
  1171. int flush = 1;
  1172. int proto;
  1173. off = skb_gro_offset(skb);
  1174. hlen = off + sizeof(*iph);
  1175. iph = skb_gro_header_fast(skb, off);
  1176. if (skb_gro_header_hard(skb, hlen)) {
  1177. iph = skb_gro_header_slow(skb, hlen, off);
  1178. if (unlikely(!iph))
  1179. goto out;
  1180. }
  1181. proto = iph->protocol;
  1182. rcu_read_lock();
  1183. ops = rcu_dereference(inet_offloads[proto]);
  1184. if (!ops || !ops->callbacks.gro_receive)
  1185. goto out_unlock;
  1186. if (*(u8 *)iph != 0x45)
  1187. goto out_unlock;
  1188. if (unlikely(ip_fast_csum((u8 *)iph, 5)))
  1189. goto out_unlock;
  1190. id = ntohl(*(__be32 *)&iph->id);
  1191. flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
  1192. id >>= 16;
  1193. for (p = *head; p; p = p->next) {
  1194. struct iphdr *iph2;
  1195. if (!NAPI_GRO_CB(p)->same_flow)
  1196. continue;
  1197. iph2 = ip_hdr(p);
  1198. if ((iph->protocol ^ iph2->protocol) |
  1199. ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
  1200. ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
  1201. NAPI_GRO_CB(p)->same_flow = 0;
  1202. continue;
  1203. }
  1204. /* All fields must match except length and checksum. */
  1205. NAPI_GRO_CB(p)->flush |=
  1206. (iph->ttl ^ iph2->ttl) |
  1207. (iph->tos ^ iph2->tos) |
  1208. ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
  1209. NAPI_GRO_CB(p)->flush |= flush;
  1210. }
  1211. NAPI_GRO_CB(skb)->flush |= flush;
  1212. skb_gro_pull(skb, sizeof(*iph));
  1213. skb_set_transport_header(skb, skb_gro_offset(skb));
  1214. pp = ops->callbacks.gro_receive(head, skb);
  1215. out_unlock:
  1216. rcu_read_unlock();
  1217. out:
  1218. NAPI_GRO_CB(skb)->flush |= flush;
  1219. return pp;
  1220. }
  1221. static int inet_gro_complete(struct sk_buff *skb)
  1222. {
  1223. __be16 newlen = htons(skb->len - skb_network_offset(skb));
  1224. struct iphdr *iph = ip_hdr(skb);
  1225. const struct net_offload *ops;
  1226. int proto = iph->protocol;
  1227. int err = -ENOSYS;
  1228. csum_replace2(&iph->check, iph->tot_len, newlen);
  1229. iph->tot_len = newlen;
  1230. rcu_read_lock();
  1231. ops = rcu_dereference(inet_offloads[proto]);
  1232. if (WARN_ON(!ops || !ops->callbacks.gro_complete))
  1233. goto out_unlock;
  1234. err = ops->callbacks.gro_complete(skb);
  1235. out_unlock:
  1236. rcu_read_unlock();
  1237. return err;
  1238. }
  1239. int inet_ctl_sock_create(struct sock **sk, unsigned short family,
  1240. unsigned short type, unsigned char protocol,
  1241. struct net *net)
  1242. {
  1243. struct socket *sock;
  1244. int rc = sock_create_kern(family, type, protocol, &sock);
  1245. if (rc == 0) {
  1246. *sk = sock->sk;
  1247. (*sk)->sk_allocation = GFP_ATOMIC;
  1248. /*
  1249. * Unhash it so that IP input processing does not even see it,
  1250. * we do not wish this socket to see incoming packets.
  1251. */
  1252. (*sk)->sk_prot->unhash(*sk);
  1253. sk_change_net(*sk, net);
  1254. }
  1255. return rc;
  1256. }
  1257. EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
  1258. unsigned long snmp_fold_field(void __percpu *mib[], int offt)
  1259. {
  1260. unsigned long res = 0;
  1261. int i, j;
  1262. for_each_possible_cpu(i) {
  1263. for (j = 0; j < SNMP_ARRAY_SZ; j++)
  1264. res += *(((unsigned long *) per_cpu_ptr(mib[j], i)) + offt);
  1265. }
  1266. return res;
  1267. }
  1268. EXPORT_SYMBOL_GPL(snmp_fold_field);
  1269. #if BITS_PER_LONG==32
  1270. u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
  1271. {
  1272. u64 res = 0;
  1273. int cpu;
  1274. for_each_possible_cpu(cpu) {
  1275. void *bhptr;
  1276. struct u64_stats_sync *syncp;
  1277. u64 v;
  1278. unsigned int start;
  1279. bhptr = per_cpu_ptr(mib[0], cpu);
  1280. syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
  1281. do {
  1282. start = u64_stats_fetch_begin_bh(syncp);
  1283. v = *(((u64 *) bhptr) + offt);
  1284. } while (u64_stats_fetch_retry_bh(syncp, start));
  1285. res += v;
  1286. }
  1287. return res;
  1288. }
  1289. EXPORT_SYMBOL_GPL(snmp_fold_field64);
  1290. #endif
  1291. int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
  1292. {
  1293. BUG_ON(ptr == NULL);
  1294. ptr[0] = __alloc_percpu(mibsize, align);
  1295. if (!ptr[0])
  1296. return -ENOMEM;
  1297. #if SNMP_ARRAY_SZ == 2
  1298. ptr[1] = __alloc_percpu(mibsize, align);
  1299. if (!ptr[1]) {
  1300. free_percpu(ptr[0]);
  1301. ptr[0] = NULL;
  1302. return -ENOMEM;
  1303. }
  1304. #endif
  1305. return 0;
  1306. }
  1307. EXPORT_SYMBOL_GPL(snmp_mib_init);
  1308. void snmp_mib_free(void __percpu *ptr[SNMP_ARRAY_SZ])
  1309. {
  1310. int i;
  1311. BUG_ON(ptr == NULL);
  1312. for (i = 0; i < SNMP_ARRAY_SZ; i++) {
  1313. free_percpu(ptr[i]);
  1314. ptr[i] = NULL;
  1315. }
  1316. }
  1317. EXPORT_SYMBOL_GPL(snmp_mib_free);
  1318. #ifdef CONFIG_IP_MULTICAST
  1319. static const struct net_protocol igmp_protocol = {
  1320. .handler = igmp_rcv,
  1321. .netns_ok = 1,
  1322. };
  1323. #endif
  1324. static const struct net_protocol tcp_protocol = {
  1325. .early_demux = tcp_v4_early_demux,
  1326. .handler = tcp_v4_rcv,
  1327. .err_handler = tcp_v4_err,
  1328. .no_policy = 1,
  1329. .netns_ok = 1,
  1330. };
  1331. static const struct net_offload tcp_offload = {
  1332. .callbacks = {
  1333. .gso_send_check = tcp_v4_gso_send_check,
  1334. .gso_segment = tcp_tso_segment,
  1335. .gro_receive = tcp4_gro_receive,
  1336. .gro_complete = tcp4_gro_complete,
  1337. },
  1338. };
  1339. static const struct net_protocol udp_protocol = {
  1340. .handler = udp_rcv,
  1341. .err_handler = udp_err,
  1342. .no_policy = 1,
  1343. .netns_ok = 1,
  1344. };
  1345. static const struct net_offload udp_offload = {
  1346. .callbacks = {
  1347. .gso_send_check = udp4_ufo_send_check,
  1348. .gso_segment = udp4_ufo_fragment,
  1349. },
  1350. };
  1351. static const struct net_protocol icmp_protocol = {
  1352. .handler = icmp_rcv,
  1353. .err_handler = ping_err,
  1354. .no_policy = 1,
  1355. .netns_ok = 1,
  1356. };
  1357. static __net_init int ipv4_mib_init_net(struct net *net)
  1358. {
  1359. if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
  1360. sizeof(struct tcp_mib),
  1361. __alignof__(struct tcp_mib)) < 0)
  1362. goto err_tcp_mib;
  1363. if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
  1364. sizeof(struct ipstats_mib),
  1365. __alignof__(struct ipstats_mib)) < 0)
  1366. goto err_ip_mib;
  1367. if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
  1368. sizeof(struct linux_mib),
  1369. __alignof__(struct linux_mib)) < 0)
  1370. goto err_net_mib;
  1371. if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
  1372. sizeof(struct udp_mib),
  1373. __alignof__(struct udp_mib)) < 0)
  1374. goto err_udp_mib;
  1375. if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
  1376. sizeof(struct udp_mib),
  1377. __alignof__(struct udp_mib)) < 0)
  1378. goto err_udplite_mib;
  1379. if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
  1380. sizeof(struct icmp_mib),
  1381. __alignof__(struct icmp_mib)) < 0)
  1382. goto err_icmp_mib;
  1383. net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
  1384. GFP_KERNEL);
  1385. if (!net->mib.icmpmsg_statistics)
  1386. goto err_icmpmsg_mib;
  1387. tcp_mib_init(net);
  1388. return 0;
  1389. err_icmpmsg_mib:
  1390. snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
  1391. err_icmp_mib:
  1392. snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
  1393. err_udplite_mib:
  1394. snmp_mib_free((void __percpu **)net->mib.udp_statistics);
  1395. err_udp_mib:
  1396. snmp_mib_free((void __percpu **)net->mib.net_statistics);
  1397. err_net_mib:
  1398. snmp_mib_free((void __percpu **)net->mib.ip_statistics);
  1399. err_ip_mib:
  1400. snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
  1401. err_tcp_mib:
  1402. return -ENOMEM;
  1403. }
  1404. static __net_exit void ipv4_mib_exit_net(struct net *net)
  1405. {
  1406. kfree(net->mib.icmpmsg_statistics);
  1407. snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
  1408. snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
  1409. snmp_mib_free((void __percpu **)net->mib.udp_statistics);
  1410. snmp_mib_free((void __percpu **)net->mib.net_statistics);
  1411. snmp_mib_free((void __percpu **)net->mib.ip_statistics);
  1412. snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
  1413. }
  1414. static __net_initdata struct pernet_operations ipv4_mib_ops = {
  1415. .init = ipv4_mib_init_net,
  1416. .exit = ipv4_mib_exit_net,
  1417. };
  1418. static int __init init_ipv4_mibs(void)
  1419. {
  1420. return register_pernet_subsys(&ipv4_mib_ops);
  1421. }
  1422. static int ipv4_proc_init(void);
  1423. /*
  1424. * IP protocol layer initialiser
  1425. */
  1426. static struct packet_offload ip_packet_offload __read_mostly = {
  1427. .type = cpu_to_be16(ETH_P_IP),
  1428. .callbacks = {
  1429. .gso_send_check = inet_gso_send_check,
  1430. .gso_segment = inet_gso_segment,
  1431. .gro_receive = inet_gro_receive,
  1432. .gro_complete = inet_gro_complete,
  1433. },
  1434. };
  1435. static int __init ipv4_offload_init(void)
  1436. {
  1437. /*
  1438. * Add offloads
  1439. */
  1440. if (inet_add_offload(&udp_offload, IPPROTO_UDP) < 0)
  1441. pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
  1442. if (inet_add_offload(&tcp_offload, IPPROTO_TCP) < 0)
  1443. pr_crit("%s: Cannot add TCP protocol offlaod\n", __func__);
  1444. dev_add_offload(&ip_packet_offload);
  1445. return 0;
  1446. }
  1447. fs_initcall(ipv4_offload_init);
  1448. static struct packet_type ip_packet_type __read_mostly = {
  1449. .type = cpu_to_be16(ETH_P_IP),
  1450. .func = ip_rcv,
  1451. };
  1452. static int __init inet_init(void)
  1453. {
  1454. struct inet_protosw *q;
  1455. struct list_head *r;
  1456. int rc = -EINVAL;
  1457. BUILD_BUG_ON(sizeof(struct inet_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
  1458. sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
  1459. if (!sysctl_local_reserved_ports)
  1460. goto out;
  1461. rc = proto_register(&tcp_prot, 1);
  1462. if (rc)
  1463. goto out_free_reserved_ports;
  1464. rc = proto_register(&udp_prot, 1);
  1465. if (rc)
  1466. goto out_unregister_tcp_proto;
  1467. rc = proto_register(&raw_prot, 1);
  1468. if (rc)
  1469. goto out_unregister_udp_proto;
  1470. rc = proto_register(&ping_prot, 1);
  1471. if (rc)
  1472. goto out_unregister_raw_proto;
  1473. /*
  1474. * Tell SOCKET that we are alive...
  1475. */
  1476. (void)sock_register(&inet_family_ops);
  1477. #ifdef CONFIG_SYSCTL
  1478. ip_static_sysctl_init();
  1479. #endif
  1480. tcp_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
  1481. /*
  1482. * Add all the base protocols.
  1483. */
  1484. if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
  1485. pr_crit("%s: Cannot add ICMP protocol\n", __func__);
  1486. if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
  1487. pr_crit("%s: Cannot add UDP protocol\n", __func__);
  1488. if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
  1489. pr_crit("%s: Cannot add TCP protocol\n", __func__);
  1490. #ifdef CONFIG_IP_MULTICAST
  1491. if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
  1492. pr_crit("%s: Cannot add IGMP protocol\n", __func__);
  1493. #endif
  1494. /* Register the socket-side information for inet_create. */
  1495. for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
  1496. INIT_LIST_HEAD(r);
  1497. for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
  1498. inet_register_protosw(q);
  1499. /*
  1500. * Set the ARP module up
  1501. */
  1502. arp_init();
  1503. /*
  1504. * Set the IP module up
  1505. */
  1506. ip_init();
  1507. tcp_v4_init();
  1508. /* Setup TCP slab cache for open requests. */
  1509. tcp_init();
  1510. /* Setup UDP memory threshold */
  1511. udp_init();
  1512. /* Add UDP-Lite (RFC 3828) */
  1513. udplite4_register();
  1514. ping_init();
  1515. /*
  1516. * Set the ICMP layer up
  1517. */
  1518. if (icmp_init() < 0)
  1519. panic("Failed to create the ICMP control socket.\n");
  1520. /*
  1521. * Initialise the multicast router
  1522. */
  1523. #if defined(CONFIG_IP_MROUTE)
  1524. if (ip_mr_init())
  1525. pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
  1526. #endif
  1527. /*
  1528. * Initialise per-cpu ipv4 mibs
  1529. */
  1530. if (init_ipv4_mibs())
  1531. pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
  1532. ipv4_proc_init();
  1533. ipfrag_init();
  1534. dev_add_pack(&ip_packet_type);
  1535. rc = 0;
  1536. out:
  1537. return rc;
  1538. out_unregister_raw_proto:
  1539. proto_unregister(&raw_prot);
  1540. out_unregister_udp_proto:
  1541. proto_unregister(&udp_prot);
  1542. out_unregister_tcp_proto:
  1543. proto_unregister(&tcp_prot);
  1544. out_free_reserved_ports:
  1545. kfree(sysctl_local_reserved_ports);
  1546. goto out;
  1547. }
  1548. fs_initcall(inet_init);
  1549. /* ------------------------------------------------------------------------ */
  1550. #ifdef CONFIG_PROC_FS
  1551. static int __init ipv4_proc_init(void)
  1552. {
  1553. int rc = 0;
  1554. if (raw_proc_init())
  1555. goto out_raw;
  1556. if (tcp4_proc_init())
  1557. goto out_tcp;
  1558. if (udp4_proc_init())
  1559. goto out_udp;
  1560. if (ping_proc_init())
  1561. goto out_ping;
  1562. if (ip_misc_proc_init())
  1563. goto out_misc;
  1564. out:
  1565. return rc;
  1566. out_misc:
  1567. ping_proc_exit();
  1568. out_ping:
  1569. udp4_proc_exit();
  1570. out_udp:
  1571. tcp4_proc_exit();
  1572. out_tcp:
  1573. raw_proc_exit();
  1574. out_raw:
  1575. rc = -ENOMEM;
  1576. goto out;
  1577. }
  1578. #else /* CONFIG_PROC_FS */
  1579. static int __init ipv4_proc_init(void)
  1580. {
  1581. return 0;
  1582. }
  1583. #endif /* CONFIG_PROC_FS */
  1584. MODULE_ALIAS_NETPROTO(PF_INET);